Approximate solutions in multiobjective interval-valued optimization problems: Existence theorems and optimality conditions

Chuang-liang Zhang, Yun-cheng Liu, Nan-jing Huang · 2025

Read Approximate solutions in multiobjective interval-valued optimization problems: Existence theorems and optimality conditions by Chuang-liang Zhang, Yun-cheng Liu, Nan-jing Huang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-12506.txt.

Introduction

Chuang-liang Zhang, Yun-cheng Liu, and Nan-jing Huang study approximate solutions of optimization problems where objectives come as intervals rather than exact numbers, in a two thousand twenty-five math paper. The setting is multiobjective, so trade-offs between goals matter as well.

Working from an interval order relation, the authors prove new existence theorems for approximate solutions under mild settings. They then derive KKT optimality settings suited to nonsmooth, nonconvex objective functions, and propose a useful approximate KKT setting. The machinery is applied to a noncooperative game whose payoffs are multiobjective interval-valued functions, showing the theory can handle strategic settings with uncertain payoffs.

Catalog record

CREATORChuang-liang Zhang, Yun-cheng Liu, Nan-jing Huang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-12506

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Approximate solutions in multiobjective interval-valued optimization problems: Existence theorems and optimality conditions Thanks: This work was supported by the National Natural Science Foundation of China (12201251,12171339), the Natural Science Foundation of Sichuan Province (2022NSFSC1830) and Southwest Minzu University Research Startup Funds (RQD2022035). Chuang-liang Zhang, Yun-cheng Liu, Nan-jing Huang Abstract Abstract . This paper is devoted to the study of approximate solutions for a multiobjective interval-valued optimization problem based on an interval order. We establish new existence theorems of approximate solutions for such a problem under some mild conditions. Moreover, we give KKT optimality conditions for approximate solutions for such a problem whose associated functions are nonsmooth and nonconvex. We also propose the approximate KKT optimality condition of an approximate solution for such a problem. Finally, we apply some obtained results to a noncooperative game involving the multiobjective interval-valued function. Keywords: Approximate solution; multiobjective optimization; interval-valued function; KKT optimality condition; approximate KKT point 2020 Mathematics Subject Classification: 90C29; 90C46; 49J52 1 Introduction Multiobjective optimization has been widely used to deal with many problems stemmed from engineering, games theory, and biology (see [ 16 , 29 ] ). We notice that numerical algorithms for solving the multiobjective optimization problems (for short, MOPs) are related to approximate solutions. Approximate solutions of MOPs can be considered as feasible points whose objective values are near to optimal values. About studying approximate solutions for MOPs, there are numerous approaches. It is worth mentioning that Loridan [ 18 ] introduced and studied \varepsilon -quasi minimal solutions for MOPs based on Ekeland’s variational principle [ 9 ] (for short, EVP). In past decades, various types of approximate solutions of MOPs have been investigated by many authors (see [ 2 , 10 , 11 , 12 , 13 , 22 ] ). In many real-life situations, the multiobjective decision-making model is often embedded with uncertain parameters that makes it difficult to obtain precise date. As is well known, interval-valued methods provide a powerful tool to deal with optimization models involving uncertain data. There have been numerous studies addressing the interval-valued optimization/analysis theory with applications (see [ 1 , 19 , 24 , 25 , 32 ] ). Recently, Zhang and Huang [ 35 ] extended the classical results for EVP to interval cases and provided some applications; Zhang et al. [ 36 ] presented interval-valued variational methods by using some results of [ 35 ] , and gave some theoretical applications concerning interval-valued functions; Villanueva et al. [ 34 ] obtained some optimality conditions for interval-valued optimization problems (for short, IOPs) based on the gH-directional derivative. Very recently, Sahu et al. [ 21 ] investigated efficient solutions for a class of vector optimization problems on an extended interval vector space and provided an application to portfolio optimization. It is worth mentioning that there has been increasing and high demanding for investigating multiobjective interval-valued optimization problems (for short, MIOPs), from theoretical, numerical and practical aspects, see, for example [ 6 , 15 , 14 , 20 , 28 , 26 , 27 , 31 , 32 , 33 ] , and the references therein. Recently, Tuyen [ 30 ] studied some existence results and KKT optimality conditions for approximate solutions of IOPs involving convexity; Zhang and Huang [ 35 ] obtained the existence of approximate solutions for IOPs by using interval EVP. However, to the best of our knowledge, existence theorems and optimality conditions for approximate solutions of MIOPs have not been considered in the literature, apart from [ 15 ] wherein some KKT-type optimality conditions were given for approximate quasi Pareto solutions of MIOPs by employing the Mordukhovich/limiting subdifferential. Thus, it would be important to investigate existence theorems and optimality conditions for approximate solutions of MIOPs under some mild conditions. The current paper is devoted to the study of approximate solutions of MIOPs. The main contributions of this work focus on some points as following: (i) a new multiobjective interval-valued version of EVP is established; (ii) new existence theorems for approximate solutions of MIOP are given; (iii) by using the nonsmooth analysis technique, new results on KKT optimality conditions for approximate solutions of MIOP are proposed under nonconvex and nonsmooth conditions. Moreover, modified \epsilon -KKT points for the weak minimal and approximate solution of such problems are presented; (iv) new results on KKT optimality conditions for approximate Nash equilibrium of a game with the multiobjective interval-value case are derived. The rest of this paper is organized as follows. Section 2 contains some basic definitions and preliminary results. Section 3 shows new results of existence of approximate solutions for MIOP. Section 4 presents KKT optimality conditions and the approximate KKT condition for approximate solutions of MIOP. In Section 5 , we present an application to a noncooperative game for multiobjective interval-valued functions. To finish, Section 6 gives some conclusions and our future research works. 2 Preliminaries Let \mathcal{I} be the set of all closed and bounded intervals in \mathbb{R} , namely, \mathcal{I}=\{[\underline{q},\overline{q}]:\underline{q}\leq\overline{q},\underline{q},\overline{q}\in\mathbb{R}\} . Let \mathbb{R}^{m}_{+} be the nonnegative orthant of \mathbb{R}^{m} . If Q\subset\mathbb{R}^{n} , then \textrm{co}Q (resp., \textrm{cl}Q ) is the convex hull (resp., the closure) of Q . Letting Q=[\underline{q},\overline{q}],R=[\underline{r},\overline{r}]\in\mathcal{I} with \alpha\in\mathbb{R} , we define Q+R=[\underline{q}+\underline{r},\overline{q}+\overline{r}]\\ and \alpha Q=\begin{cases}[\alpha\underline{q},\alpha\overline{q}]&\text{if $\alpha\geq 0,$}\\ [\alpha\overline{q},\alpha\underline{q}]&\text{if $\alpha<0.$}\end{cases} For Q=[\underline{q},\overline{q}]\in\mathcal{I} , the center and width are defined as q^{c}=\frac{\underline{q}+\overline{q}}{2} and q^{w}=\frac{\overline{q}-\underline{q}}{2} . Clearly, \underline{q}=q^{c}-q^{w} and \overline{q}=q^{c}+q^{w} . Also, the norm \|\cdot\|_{\mathcal{I}} on \mathcal{I} can be defined as [ 19 ] : \|Q\|_{\mathcal{I}}=\max\{|\underline{q}|,|\overline{q}|\} . Recall the notion of generalized Hukuhara difference (for short, gH) on \mathcal{I} as follows. Definition 2.1 . ( [ 1 , 24 ] ) Let Q,R\in\mathcal{I} . The gH between Q and R is defined as Q\ominus_{gH}R=P\Leftrightarrow Q=R+P\;\textrm{or}\;R=Q-P, where P\in\mathcal{I} . Obviously, if Q=[\underline{q},\overline{q}] and R=[\underline{r},\overline{r}] , then Q\ominus_{gH}R=\left[\min\{\underline{q}-\underline{r},\overline{q}-\overline{r}\},\max\{\underline{q}-\underline{r},\overline{q}-\overline{r}\}\right] . The Hausdorff distance h on \mathcal{I} is given by h(Q,R)=\max\left\{\sup_{q\in Q}d_{R}(q),\sup_{r\in R}d_{Q}(r)\right\},\quad\forall Q,R\in\mathcal{I}, where d_{R}(q):=\inf_{r\in R}|q-r| . Furthermore, if Q=[\underline{q},\overline{q}] and R=[\underline{r},\overline{r}] , then h(Q,R)=\|Q\ominus_{gH}R\|_{\mathcal{I}}=\max\{|\underline{q}-\underline{r}|,|\overline{q}-\overline{r}|\}. Clearly, (\mathcal{I},h) is a complete metric space. A sequence \{Q_{n}\}\subset\mathcal{I} is said to be convergence to Q\in\mathcal{I} if h(Q_{n},Q)\rightarrow 0 . Obviously, if Q_{n}=[\underline{q}_{n},\overline{q}_{n}] and Q=[\underline{q},\overline{q}] , then h(Q_{n},Q)\rightarrow 0 iff |\underline{q}_{n}-\underline{q}|\rightarrow 0 and |\overline{q}_{n}-\overline{q}|\rightarrow 0 . We introduce two interval orders on \mathcal{I} as follows. Definition 2.2 . ( [ 14 ] ) Let Q=[\underline{q},\overline{q}] and R=[\underline{r},\overline{r}] . The interval orders \preccurlyeq_{CW} and \prec_{CW} on \mathcal{I} are defined by Q\preccurlyeq_{CW}R\Leftrightarrow q^{c}\leq r^{c}\;\textrm{and}\;q^{w}\leq r^{w};\quad Q\prec_{CW}R\Leftrightarrow\ q^{c}<r^{c}\;\textrm{and}\;q^{w}<r^{w}. Remark 2.1 . Clearly, (i) \preccurlyeq_{CW} is the partial order, i.e., it is reflexive, antisymmetric and transitive; (ii) for each P\in\mathcal{I} , Q\preccurlyeq_{CW}(\textrm{resp}.,\prec_{CW})~R\Leftrightarrow Q+P\preccurlyeq_{CW}(\textrm{resp}.,\prec_{CW})~R+P ; (iii) if Q_{1}\preccurlyeq_{CW}(\textrm{resp}.,\prec_{CW})~R_{1} and Q_{2}\preccurlyeq_{CW}(\textrm{resp}.,\prec_{CW})~R_{2} for Q_{1},Q_{2},R_{1},R_{2}\in\mathcal{I} , then Q_{1}+Q_{2}\preccurlyeq_{CW}(\textrm{resp}.,\prec_{CW})~R_{1}+R_{2} ; (iv) for each \alpha\geq 0 , \alpha Q\preccurlyeq_{CW}\alpha R . Let f:\mathbb{R}^{n}\to\mathcal{I} be an interval-valued function (for short, IVF) denoted by f(u)=[\underline{f}(u),\overline{f}(u)] with \underline{f}(u)\leq\overline{f}(u) . In this case, we write f^{c}(u)=\frac{\underline{f}(u)+\overline{f}(u)}{2} and f^{w}(u)=\frac{\overline{f}(u)-\underline{f}(u)}{2} . Recall some concepts of IVFs and their properties as follows. Definition 2.3 . An IVF f:\mathbb{R}^{n}\to\mathcal{I} is said to be (i) \preccurlyeq_{CW} -lsc ( \preccurlyeq_{CW} -lower semicontinuous) on \mathbb{R}^{n} if f^{c} and f^{w} are lsc on \mathbb{R}^{n} ; (ii) [ 32 ] continuous on \mathbb{R}^{n} if for every u_{0}\in\mathbb{R}^{n} and each \{u_{n}\}\subset\mathbb{R}^{n} satisfying u_{n}\rightarrow u_{0} , one has \lim_{n\rightarrow+\infty}h(f(u_{n}),f(u_{0}))=0; (iii) locally Lipschitz continuous on \mathbb{R}^{n} if for every u_{0}\in\mathbb{R}^{n} , there exist a neighbourhood U(u_{0}) and some constant L>0

Read the complete text